Application of Neural Network Enhanced Ground-Penetrating Radar to Localization of Burial Sites

Abstract

The problem of searching for burial sites is an important issue for criminology, history, and archeology. The presently employed classical ground-penetrating radar (GPR) methods often yield equivocal results. Here, we report the results of our experimental study on the possible enhancement of the GPR methodology by introduction of the neural network to help localize the places of deposition of cadavers. The experiments, employing pig bodies as a model, yielded very promising results, but also indicated the problems that must be taken into account in further development of the methodology. Such problems include the influence of several circumstances on GPR signal and its computer-aided interpretation. Particularly important are: 1) the progressive decomposition of the body; 2) the dressing on the body—none, summer or winter garment, or wrapped in shroud or plastic foil; and 3) the size of the body. Based on the positive results of the preliminary survey the problems will be investigated using the prepared beforehand area with four burial sites and developed method for neural networks.